基于网络关系模型的品牌女装配色解析

针对当前服饰用色规律及其内在机制解析模糊的问题,提出了利用图像处理技术构建服饰配色关系及其内在联系的网络关系模型。对服饰系列图像色彩进行初始聚类及二次聚类,提取主色及其用色比例关系;提取色彩占比、二元色组共现频率、色彩空间距离等指标构建主色网络关系模型;对选取的3个代表性品牌女装的500幅图像,利用所开发的色彩解析软件,从主色分布、配色关系、配色机制方面进行了比对和分析,以可视化形式呈现品牌女装年度、时尚季用色规律。实验结果表明,3个女装品牌主色、空间距离和共现规律不尽相同,前3种颜色累计占比均超出50%,单幅样本耗时小于0.3 s;该方法能够以客观、可视化的方式表征服饰用色机制,用以辅助服装企划和设计决策。

Color parsing of female brand clothing based on nexus network modeling

Targeting at lacks of the deep level parsing on rules and mechanisms of clothing color, a nexus network model for clothing color relations and internal connection using imaging processing technology was established. The dominant colors and color ratios of female brand clothing were detected using two-step clustering before an innovative nexus network model was constructed and composed multi-features such as diff-color frequency, dual color matching intimacy and color distances. 500 female clothing images of three representative brands were chosen and analyzed in terms of main color distribution, matching relationship and mechanisms. The developed software efficiently displays each brand color pattern for different years and quarters with a visualized model. Results show that the main colors, spatial distance and co-occurrence rules are different, the cumulative proportion of the first three colors is more than 50%, and the time cost of a single image is less than 0.3 s.The visualized color nexus network model is able to display objectively clothing color rules,assisting the fashion merchandising and design decision-making.